Operating a distributed IaaS Cloud for BaBar MC production and user analysis
Presented by Ian GABLE on 5 May 2011 from 11:15 to 11:45
Session: Cloud, grid and virtualization
Track: Grid, cloud and virtualization
In the last year we have established a system which replicates a standard Condor HTC environment across multiple distinct IaaS clouds of different types including EC2, Nimbus and Eucalyptus. Users simply submit batch jobs to a Condor queue containing a custom attribute which is a pointer to the Virtual Machine Image they would like booted to service their job. The system automatically boots instances of the requested machine type on one of the available clouds and contextualizes them to connect to the batch system. The system is being used on a continual basis for astronomy and HEP jobs. We report on our experience operating this system which has booted over 30 000 VMs and completed over 250 000 jobs.
Location: GSI Helmholtzzentrum für Schwerionenforschung GmbH
Address: Planckstr. 1, 64291 Darmstadt, Germany
Room: Hörsaal / lecture hall